[PATCH v2] drm/amdkfd: not restore userptr buffer if kfd process has been removed

Chen, Xiaogang xiaogang.chen at amd.com
Thu Oct 3 22:11:14 UTC 2024


On 10/3/2024 4:11 PM, Felix Kuehling wrote:
>
> On 2024-10-03 16:55, Xiaogang.Chen wrote:
>> From: Xiaogang Chen <xiaogang.chen at amd.com>
>>
>> When kfd process has been terminated not restore userptr buffer after 
>> mmu notifier invalidates a range.
>>
>> Signed-off-by: Xiaogang Chen<Xiaogang.Chen at amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 12 ++++++++----
>>   1 file changed, 8 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> index ce5ca304dba9..1df0926b63b3 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
>> @@ -2524,11 +2524,15 @@ int amdgpu_amdkfd_evict_userptr(struct 
>> mmu_interval_notifier *mni,
>>           /* First eviction, stop the queues */
>>           r = kgd2kfd_quiesce_mm(mni->mm,
>>                          KFD_QUEUE_EVICTION_TRIGGER_USERPTR);
>> -        if (r)
>> +
>> +        if (r && r != -ESRCH)
>>               pr_err("Failed to quiesce KFD\n");
>> -        queue_delayed_work(system_freezable_wq,
>> -            &process_info->restore_userptr_work,
>> -            msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
>> +
>> +        if (!r || r != -ESRCH) {
>
> This condition is always true.
>
so sure why this condition is always true?  kgd2kfd_quiesce_mm can 
return -ESRCH when it cannot find kfd process correspondent to mni->mm, 
then above checking will be false, then will not queue restore work item 
into system_freezable_wq.

Regards

Xiaogang

> Regards,
>   Felix
>
>
>> + queue_delayed_work(system_freezable_wq,
>> +                &process_info->restore_userptr_work,
>> + msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS));
>> +        }
>>       }
>>       mutex_unlock(&process_info->notifier_lock);


More information about the amd-gfx mailing list